> ## Documentation Index > Fetch the complete documentation index at: https://docs.quiva.ai/llms.txt > Use this file to discover all available pages before exploring further. # Creating Your First Assistant > Build and deploy your first AI assistant in minutes This guide walks you through creating your first AI assistant from scratch. By the end, you'll have a configured assistant ready to use — and optionally shared with your team. **Time to complete**: 5-10 minutes\ **What you'll build**: A custom AI assistant configured for your use case\ **What you'll learn**: The three creation paths, configuration basics, and testing in chat ## Prerequisites Sign up at [app.quiva.ai](https://app.quiva.ai/en/signup) Check your inbox and verify your email address ## Three Ways to Create an Assistant QuivaWorks offers three paths to create an assistant: Describe what you need and QuivaWorks builds the initial configuration for you. Best for starting from scratch. Duplicate one of your existing assistants as a starting point. Best when you have something similar already. Install a pre-built assistant from the community. Best for common use cases with a head start. In this guide, we'll use **Create with AI** — the fastest path for most people. *** ## Step 1: Start Creation 1. In the left sidebar, navigate to **Assistants** 2. Click the **Create Assistant** button 3. Select **Create with AI** A setup dialog will appear. Describe what you want your assistant to do: ``` A customer service assistant for a SaaS product. It should answer questions about the product, help users troubleshoot issues, and know when to escalate to a human agent. ``` Click **Generate** and QuivaWorks will create an initial configuration with a suggested name, description, and instructions. Be specific in your description. Mentioning the domain, the task, and any constraints (like "escalate when needed") produces much better initial configurations. *** ## Step 2: Review Information Settings After generation, you'll land in the **Information** tab. Review and refine what was generated. ### Name & Description Give your assistant a clear name that team members will recognise: ``` Name: Customer Support Description: Handles product questions, troubleshooting, and support escalations ``` ### Team vs. Personal Choose who can access this assistant: Visible to all members of your account. Use for shared workflows, standardised processes, and company-wide tools. **Choose this for most production assistants.** Private to you only. Use for experiments, individual workflows, or personal use cases. ### Instructions Instructions define your assistant's role, personality, and behaviour. They're the most impactful part of your configuration: ``` You are a customer support specialist for [Your Product]. Your responsibilities: - Answer questions about product features and pricing - Help users troubleshoot common issues step-by-step - Escalate to a human agent when needed: billing disputes, account security issues, or when a user explicitly asks for a person Your personality: - Friendly, professional, and patient - Solution-focused — always try to resolve the issue before escalating - Concise — don't over-explain unless asked Guidelines: - If you don't know something, say so and offer to connect them with the team - For billing or account issues, escalate to a human - Always confirm the issue is resolved before ending the conversation ``` Good instructions include: role, responsibilities, personality, tone, and specific behavioural guidelines. Think of it as onboarding a new team member in writing. *** ## Step 3: Configure Provider Settings Click the **Provider** tab to set the AI model. All assistants use **Claude Haiku 4.5** by default — fast, cost-effective, and included in all QuivaWorks plans. No API key required. Your plan's credits are used automatically. Claude Haiku 4.5 handles customer service, research, content, data extraction, and the vast majority of business workflows extremely well. It's the right starting point for almost every assistant. On **Team and Enterprise plans**, you can connect your own Anthropic API key to use specific Claude models. When using your own key, you're billed directly by Anthropic instead of consuming QuivaWorks credits. [Learn more →](/assistants/configuration/provider-settings) For this guide, leave the default selected and move on. *** ## Step 4: Add Knowledge (Optional) Click the **Knowledge** tab to give your assistant access to your specific content. Knowledge lets your assistant answer questions based on your documentation, FAQs, policies, or any other reference material — not just its general training. You can add knowledge from: * **File upload** — PDFs, Word documents, text files * **URL** — Web pages and documentation sites * **Manual entry** — Paste text directly For your first assistant, you can skip this and add knowledge later. The assistant will still work based on its instructions alone. How to add and manage knowledge sources → *** ## Step 5: Connect Integrations (Optional) Click the **Integrations** tab to connect your assistant to external systems via MCP. Integrations give assistants the ability to take actions — look up orders, search a CRM, create tickets, or call your APIs. For your first assistant, skip this for now and add integrations as you need them. How to connect tools and APIs → *** ## Step 6: Save and Open in Chat Click **Save** to save your configuration, then click **Open in Chat** to test it. Try a few questions that match your use case: * "What features does your product have?" * "I'm having trouble logging in, can you help?" * "I want to speak to a human" Evaluate the responses: are they on-brand? Does the assistant stay in scope? Do edge cases (like asking for a human) behave correctly? If something is off, go back to **Instructions** and refine the guidance — then test again. Use the thumbs up/down buttons on responses to flag what works and what doesn't. These votes feed into the [Learning system](/assistants/learning) so you can improve the assistant over time. *** ## Step 7: Share with Your Team (Optional) If you created a **Team Assistant**, your team can already find and use it in their own sessions. To collaborate on a specific conversation in real-time: 1. Open the assistant in chat 2. Type `@` to mention a team member 3. They'll receive a notification and can join the session instantly Real-time session sharing and @mentions → *** ## What's Next Your assistant is live. Here are the most impactful ways to build on it: Give your assistant access to your documentation and content Link to your CRM, ticketing system, or custom APIs Use feedback votes and insights to refine behaviour over time Trigger your assistant automatically via webhook, schedule, or email Write better instructions for better results Explore all configuration options in depth